EEVblog #474 - GW Instek GDS-2000A Series Oscilloscope Unboxing & Fi...

[marmad]

You set the acquisition memory into “short” mode, so the envelope of the AM signal is not solid, if you change to “Auto” mode, there will be no problem.

Are they saying that this is what caused the strange spikes you saw in the AM signal? Because of the sample rate derived from the memory length?

[EEVblog]

Are they saying that this is what caused the strange spikes you saw in the AM signal? Because of the sample rate derived from the memory length?

Not entirely sure 

[marmad]

At 43:00, you compare the AM measurement with Rigol. GDS is far behind Rigol regard to the Intensity performance.

Anyway, kudos to them for at least saying this straight up.

[vk6zgo]

Are they saying that this is what caused the strange spikes you saw in the AM signal? Because of the sample rate derived from the memory length?

Not entirely sure 

Dave,I was going to point out that an overmodulated AM signal like you used will produce harmonics of the carrier & sidebands,as well as intermodulation products between these harmonics & the original signals.

I thought maybe the GWInstek was having problems with aliasing with these signals.
If so,it's a sampling rate problem,but wouldn't happen with a lower % of modulation.

[madires]

Honestly, you don't want the FG as it is right now.  That's not to say they won't release a new FG module later but the current one - meh - for the same price get a dedicated device with a heck of a lot more functionality.

Yep, for the same money (€249) I could buy a Siglent SDG1010. And if you consider that the scope already got the PSU, screen and knobs it's hard to believe how much GW Instek charges for 5MHz and 3 waveforms.

[grego]

Are they saying that this is what caused the strange spikes you saw in the AM signal? Because of the sample rate derived from the memory length?

Not entirely sure 

If you reply back to them directly they will probably work to clear it up.  They've been fantastic at getting back to me.  Copy the guy who sent you the scope too since he's probably in marketing and wants to see positive reviews.

[Bored@Work]

They've been fantastic at getting back to me.  Copy the guy who sent you the scope too since he's probably in marketing and wants to see positive reviews.

Thats how it started when we had a problem, too.  But the Taiwanese started to duck and cover and felt silent when it became difficult because their distributors messed up.

[grego]

They've been fantastic at getting back to me.  Copy the guy who sent you the scope too since he's probably in marketing and wants to see positive reviews.

Thats how it started when we had a problem, too.  But the Taiwanese started to duck and cover and felt silent when it became difficult because their distributors messed up.

Again I deal with InstekUSA so a different beast.

[marmad]

You set the acquisition memory into “short” mode, so the envelope of the AM signal is not solid, if you change to “Auto” mode, there will be no problem.

Well, assuming the Instek follows the normal math, 'Short' mode means 1M record length - and you were on 200us, so it was likely sampling at 500MSa/a - but you had the RIgol set to AUTO so it was sampling at 2GSa/s - so I guess it wasn't a fair comparison. Perhaps check it again with the same signal on AUTO and sampling @ 2GSa/s?

[mike1305]

You set the acquisition memory into “short” mode, so the envelope of the AM signal is not solid, if you change to “Auto” mode, there will be no problem.

Well, assuming the Instek follows the normal math, 'Short' mode means 1M record length - and you were on 200us, so it was likely sampling at 500MSa/a - but you had the RIgol set to AUTO so it was sampling at 2GSa/s - so I guess it wasn't a fair comparison. Perhaps check it again with the same signal on AUTO and sampling @ 2GSa/s?

Short mode switches to 1,000 points of memory.

[Hydrawerk]

Really 1kpoints? Is this in the manual?

[grego]

You set the acquisition memory into “short” mode, so the envelope of the AM signal is not solid, if you change to “Auto” mode, there will be no problem.

Well, assuming the Instek follows the normal math, 'Short' mode means 1M record length - and you were on 200us, so it was likely sampling at 500MSa/a - but you had the RIgol set to AUTO so it was sampling at 2GSa/s - so I guess it wasn't a fair comparison. Perhaps check it again with the same signal on AUTO and sampling @ 2GSa/s?

Short mode switches to 1,000 points of memory.

Incorrect sir.  From the manual:

There are two record length settings, Auto and Short. The Auto setting will set the record length to maximum record length available, dependent on the scope settings. The Short setting will set the record length to 1M.

[mike1305]

There are two record length settings, Auto and Short. The Auto setting will set the record length to maximum record length available, dependent on the scope settings. The Short setting will set the record length to 1M.

Looks like a typo, did a quick test with short memory mode:

Set to 500 ns/div-> samples at 200 MSa/s

(500 ns/div) x (10 div) x (200 MSa/s) = (5 us)(200 MSa/s) = 1000 Sa

[marmad]

Wouldn't be surprised if the manual had a typo. It's 1,000 points, I can guarantee it. Did some testing on sample rate at various timebases. There is no way they are using 1M points of memory at 200ns/div and getting 80k wfm/s performance.

It would be more than a typo - it would be multiple typos - no where in the manual does it mention 1k. I think it's 1M (500k with 2 channels), which would be rather standard these days.

[mike1305]

Wow, I wonder where they get that from. What's the point of "short" memory that's only half the depth? Why not just turn on channel 2 to half the memory depth? 

Also, check my edit above. Perhaps Dave can verify when he does a full review/comparo, I may just be going bonkers! 

[marmad]

Wow, I wonder where they get that from. What's the point of "short" memory that's only half the depth? Why not just turn on channel 2 to half the memory depth? 

Also, check my edit above. Perhaps Dave can verify when he does a full review/comparo, I may just be going bonkers! 

500ns/div with just a single channel on has a sampling rate of 200MSa/a? Yes, you must be right - SHORT must be 1k - and AUTO must give you the 1M - and it must be a typo on the page before the table I posted. I guess it makes more sense since 'SHORT' seemed like an odd thing to call 1M.

Edit: That means in Dave's vid - if he actually had the 'SHORT' selected - he was sampling at 500kSa/s! No wonder he got the glitches!

[Hydrawerk]

And the GDS-806C from 2004 or so has this funny memory... It was sometimes 125kpoints and sometimes 0,5kpoints. 

[Yaksaredabomb]

...Perhaps Dave can verify when he does a full review/comparo....

That would be neat to see, an update and compare after some time with the manual and with Instek's comments.  Hope it works out - even if it has to be less than an hour to allow Dave to create it in a reasonable time.  (though I side with the "longer, less-edited, unscripted" crowd)

[marmad]

...Perhaps Dave can verify when he does a full review/comparo....

That would be neat to see, an update and compare after some time with the manual and with Instek's comments.  Hope it works out - even if it has to be less than an hour to allow Dave to create it in a reasonable time.  (though I side with the "longer, less-edited, unscripted" crowd)

As mentioned in the other thread, the 1k Short record length is already confirmed in the video this thread is attached to (check rate divided-by timebase between 53:20 - 53:40) - and also in the Instek's FW HELP files. I'm not sure if any further comments from Instek will be forthcoming - or from Dave is really necessary.

I wouldn't mind seeing someone measure the waveform update rates of the Instek at the AUTO (1M) settings which produce the 2GSa/s rate (not the ETS rates) - for a more realistic speed comparison against the Rigol DS2000 (for which I've posted the numbers below) - but I doubt Dave has time for that - and Greg doesn't seem to be posting much info any more.

[grego]

Greg is busy with work and travel lately.  Sorry.   Calms back down in a couple of weeks.

[Yaksaredabomb]

Greg is busy with work and travel lately.  Sorry.   Calms back down in a couple of weeks.

That's alright.  We'll be excited to have you back though!
 
More time with the LA decode functions would be another nice thing.  Seems like they could be better than nothing but not much more.  I seem to remember reading they're not nearly as nice/useful/refined as the Rigol and Agilent decode options, but don't quote me on that.
 
Edit June 6th: My display name has recently changed from "jneumann" to "Yaksaredabomb"

[grego]

As a side note, got a note back from Instek:

I have spoken to our factory engineers, and you are correct. There was an error on the user manual. The memory is 1K points. I had them update the user manual, and it has been uploaded on the website. I sincerely apologize for the error, it was not our intention to mislead the consumer. If you have any further questions, please call or email me.

Moderately annoying since it was one of the things I was looking at when I was evaluating.

[marmad]

As a side note, got a note back from Instek:

I have spoken to our factory engineers, and you are correct. There was an error on the user manual. The memory is 1K points. I had them update the user manual, and it has been uploaded on the website. I sincerely apologize for the error, it was not our intention to mislead the consumer. If you have any further questions, please call or email me.

Moderately annoying since it was one of the things I was looking at when I was evaluating.

Strange that the Help screen from the DSO is also incorrect - but with a different number (meaning for 2 channels?)

[Hydrawerk]

OK, this table is for 1kpoints per channel.

Now what is it like for Auto record length?
Definitely this scope is worse than Rigol DS2000 and Agilent DSOX2000 at waveform update rate. Now how does the table change when you turn on all channels, FFT or logic analyzer?
DS2000 must be better when it has two FPGAs.
Here is the Rigol chart. https://www.eevblog.com/forum/testgear/first-impressions-and-review-of-the-rigol-ds2072-ds2000-series-dso/?action=dlattach;attach=32606
It has up to 50 000 waveforms per second at 14kpoints per channel...

[Hydrawerk]

Here's another Marmad's table for DS2000.